Xenix Native

Xenix Native is a local desktop workbench for business data analysis through conversation, built with Python and PySide6 Qt Widgets.

Repository routing and knowledge ownership start in AGENTS.md. Use the local contributor guide for setup, feedback loops, testing, and release rules; the optional SVC corpus provides additional working-method guidance through svc lookup.

Quick Start

pdm sync --clean -G :all
pdm run prepare
pdm run dev --isolated

Use Python 3.14.2 and PDM 2.26.6. The authoritative Python and dependency constraints are declared in pyproject.toml and pdm.lock; project-local environment selection is declared in pdm.toml.

Contributor, test, translation, and packaging commands are owned by CONTRIBUTING.md.

--isolated starts with a disposable runtime home. Use pdm run dev for your normal persistent workspace. Run pdm run verify for the local static, offline acceptance, and isolated desktop smoke gates; focus a test with pdm run test tests/test_job_scheduler.py -q.

Layout

• src/xenix contains the application package, bootstrap code, UI, runtime config, storage services, logging, and exception handling.
• tests contains the offline acceptance portfolio, opt-in runtime/storage fixtures, offscreen UI contracts, native Windows smoke, and separately invoked live Agent benchmarks.
• scripts contains developer helpers used by pdm run.
• ml contains legacy model scripts; native ML implementation lives under src/xenix/services/ml.
• docs stores durable project knowledge.
• tasks stores volatile task packets, working evidence, and historical execution records; it is not a durable source of current product or architecture truth.
• xenix.spec is the canonical Windows PyInstaller onedir spec.

Documentation Model

• PRD: docs/10-prd/README.md
• Product TDD: docs/20-prd-tdd/README.md
• Unit TDD: docs/30-unit-tdd/README.md
• Deployment: docs/40-deployment/README.md
• Contributor workflow: CONTRIBUTING.md
• Application assembly and cleanup: composition design
• Data cleaning ownership: cleaning design
• SVC corpus baseline: svc.json

For packaged delivery, see Packaging. For resolved application paths, persisted state, and recovery, see Runtime State.